陆地碳源汇是生态模型研究全球碳循环的重要目标之一，如何优化其参数使得模型结果更靠近观测成为一个亟待解决的问题，由于生态模型的计算成本大，直接用搜索方法或者数值导数的方法会导致计算时间太长而没有实际意义。本文的目的为如何用尽量少的点将参数估计的更加精确。本文从试验设计的角度出发，比较正交试验布点、均匀设计布点、随机布点的优劣性，利用搜索非退化设计矩阵的方法，寻找使得设计矩阵行列式最大的布点方式，然后和响应曲面方法结合对模型进行拟合，以求得函数的最小值。在模拟中，该方法分别对二次函数、带误差函数、非二次函数、不可导函数进行验证，模拟结果显示，在最优点附近该方法基本上都得到了函数的最小值。然后将该方法应用于BEPS模型(Boreal Ecosystem Productivity Simulator)，选取中国东北部的一块针叶林区域, 优化该区域2003年第一周的生态参数, 包括光合作用参数(Vcmax)、呼吸作用参数(Q10)、碳库乘数因子(Multiplier to carbon pool)，得到三个参数的估计结果分别为：50:9553、2:3795、1:0035，得到通量的MSE的数量级为10ˆ(−6)，并且评价了该方法的估计精度，得到样本的均方根误差的数量级为10ˆ(−8)，说明该方法的精度很高。

Ecological model is an important model to estimate the terrestrial carbon sinks, it is aconsiderable problem to optimize its parameters so that the model’s results can be closer tothe observation. Due to the large computational cost of ecological models, the direct searchmethods or numerical derivative approach will lead to so long calculation time that there isno practical significance. The purpose of this paper is how to use less point and estimatemore accurate parameters. This paper based on experimental design, comparing the ad-vantages and disadvantages of orthogonal design grid method, uniform design grid method,random observation data, and use the method of search non-degenerate design matrix tofind the distribution method which makes the determinant of design matrix is maximum,and then combine with response surface method to fit the model, in order to obtain theminimum of the function. In the simulation, this paper uses quadratic function, quadraticfunction with the error ,non-quadratic function, un-derivative function to verify the method,simulation results show that this method have been essentially get the minimum value ofthe function near the optimum point. Finally, the method is used in BEPS model (BorealEcosystem Productivity Simulator), in this paper the northeast corner of China is selectedand the time is the first week of 2003, estimates the photosynthesis parameter(Vcmax),respiration parameter(Q10), carbon pool(Multiplier to carbon pool ) of coniferous forests,and the initial results show that the three parameters are 50:9553, 2:3795, 1:0035, and theMSE of flux is the order of 10ˆ(−6), then evaluate the accuracy of this method, the MSE isthe order of 10ˆ(−8). which means this method has very high precision.